Development of Recurrent Neural Networks for Thermal/Electrical Analysis of Non-Residential Buildings Based on Energy Consumptions Data

Elisa Belloni (), Flavia Forconi, Gabriele Maria Lozito, Martina Palermo, Michele Quercio and Francesco Riganti Fulginei
Additional contact information
Elisa Belloni: Department of Engineering, University of Perugia, 06125 Perugia, Italy
Flavia Forconi: Department of Industrial, Electronic and Mechanical Engineering, Roma Tre University, 00146 Rome, Italy
Gabriele Maria Lozito: Department of Information Engineering, University of Firenze, 50139 Firenze, Italy
Martina Palermo: Department of Industrial, Electronic and Mechanical Engineering, Roma Tre University, 00146 Rome, Italy
Michele Quercio: Department of Industrial, Electronic and Mechanical Engineering, Roma Tre University, 00146 Rome, Italy
Francesco Riganti Fulginei: Department of Industrial, Electronic and Mechanical Engineering, Roma Tre University, 00146 Rome, Italy

Energies, 2025, vol. 18, issue 12, 1-22

Abstract: Extensive research has focused on optimizing energy consumption in residential buildings based on indoor thermal conditions. However, modeling the energy and thermal behavior of non-residential buildings presents greater challenges due to their complex geometries and the high computational cost of detailed simulations. Simplifying input variables can enhance the applicability of artificial intelligence techniques in predicting energy and thermal performance. This study proposes a neural network-based approach to characterize the thermal–energy relationship in commercial buildings, aiming to provide an efficient and scalable solution for performance prediction. Consumptions trends for a building are generated using the EnergyPlus™ dynamic simulation software over a timespan of a year in different locations, and the data are then used to train neural network models. Uncertainty analyses are carried out to evaluate the behavior effectiveness of the artificial neural networks (ANNs) in different weather conditions, and the root mean square error (RMSE) is calculated in terms of mean air temperatures. The results show that this approach can reproduce the functional relationship between input and output data. Three different ANNs are trained for the northern, central, and southern climatic zones of Italy. The southern region’s models achieved the highest accuracy, with an RMSE below 0.5 °C; whereas the model for the northern cities was less accurate, since no specific trend in plant management was present, but it still achieved an acceptable accuracy of 1.0 °C. This approach is computationally lightweight; inference time is below 5 ms, and can be easily embedded in optimization algorithms for load dispatch or in microcontroller applications for building automation systems.

Keywords: mathematical modeling; artificial neural network; non-residential buildings; energy consumptions; optimization function; thermal conditions (search for similar items in EconPapers)
JEL-codes: Q Q0 Q4 Q40 Q41 Q42 Q43 Q47 Q48 Q49 (search for similar items in EconPapers)
Date: 2025
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.mdpi.com/1996-1073/18/12/3031/pdf (application/pdf)
https://www.mdpi.com/1996-1073/18/12/3031/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:gam:jeners:v:18:y:2025:i:12:p:3031-:d:1674136

Access Statistics for this article

Energies is currently edited by Ms. Agatha Cao

More articles in Energies from MDPI
Bibliographic data for series maintained by MDPI Indexing Manager ().